Broadhurst et al.: Modified sorting technique to mitigate the collateral mortality of Metapenaeus madeayi 
289 
(“15-min water-tray” treatment). The conventional-tray 
treatments involved leaving the water tray empty of wa- 
ter and removing the mesh-separating screen. The 35-L 
container for the riddled, unwanted school prawns was 
filled with water during the water-tray treatments but 
was left dry for the conventional-tray treatments. Two 
replicates of each treatment were completed on each of 
four days of fishing in the Clarence River and two days 
of fishing in the Hunter River. 
For each treatment, immediately after the unwanted 
school prawns were separated by the riddler into either 
a dry (conventional-tray treatments) or water-filled 
(water-tray treatments) 35-L container (Fig. 1), the 
transport dory was positioned alongside the trawler 
and 120 individuals were randomly selected and “dis- 
carded” into groups of 10 into 12 cages submerged in 
the water-filled 75-L PVC containers. Three of these 
cages were sampled immediately (termed T 0 — see be- 
low), before the remaining nine were transported to the 
closest monitoring site and attached to the 10-mm rope 
within 20 min, and without exposing any of the school 
prawns to air. During transfer, all caged school prawns 
were held in aerated water and the water quality was 
checked with an Horiba U10 meter (Horiba, Irvine, CA) 
and maintained (via exchange) at the same levels as 
that recorded at the surface of the deployment site. 
Within six hours of the first treatment deployment, 
240 school prawns were removed from the holding 
tanks by scoop nets and placed in groups of 10 into 
24 cages submerged in the water-filled 75-L PVC con- 
tainers onboard the dory (i.e., the same number of 
prawns as that for the two replicate deployments of 
each treatment). Six cages were sampled immediately 
(T 0 — see below), while the remaining 18 cages were 
transferred to the same monitoring sites as those hous- 
ing the treatment school prawns and used as balanced 
controls in each experiment. For each replicate of the 
control and treatment groups on each day of fishing, 
school prawns in three of the cages were sacrificed 
and sampled at three times: 24 (T 24 ), 72 (T 72 ), and 120 
( T 1 20 1 hours after T 0 . 
Data collected 
The following data were collected during each deploy- 
ment in both experiments: towing speed (m/s); fishing 
depth (m); duration of air exposure of the catch (min); air 
temperature (°C); percent cloud cover; and the numbers 
and weights (kg) of retained and discarded catches. The 
Horiba U10 was used to record replicate measures of 
water temperature (°C), dissolved oxygen (mg/L), and 
salinity (psu) in the water tray. An EC 350 Greenspan 
Smart Sensor (Tyco Environmental Systems, Lakewood, 
NJ) was attached to the trawl to provide replicate mea- 
sures of the conductivity (pS/cm) and the temperature 
of the river once each minute during each deployment 
(except on the first day of fishing in the Clarence River). 
An algorithm was used to convert the normalized con- 
ductivity readings to salinities. Means of these read- 
ings were used to provide a datum for the salinity and 
for the temperature at the surface and bottom for each 
deployment. 
At each sampling time, the three cages from the two 
replicates of the control and each of the four treatment 
groups from each day of fishing were removed from the 
monitoring site. The sediments from the cages were 
emptied onto a tray. The numbers of alive and dead 
school prawns were recorded, and, if possible, they were 
measured to the nearest 1-mm CL. During the Hunter 
River experiment, the binary molt status (hard or soft) 
of school prawns was also noted by the rigidity of their 
carapace. This was not done during the Clarence River 
experiment because all school prawns were clearly in- 
termolts (i.e., hard). For one randomly selected fishing 
day during the Clarence River experiment, two live 
school prawns were immediately selected from two of 
the cages for each replicate of the treatment and con- 
trol groups at the T 0 , T 72 , and T 120 sampling times and 
secured in aluminum satchels before being placed in 
liquid nitrogen. These frozen samples were later ana- 
lyzed for j -lactate (pmol/g) to provide an indication of 
the severity of anaerobic stress and subsequent recov- 
ery, by following the methods described by Broadhurst 
et al. (2002). 
Statistical analyses 
The data collected from each experiment were analyzed 
separately. Appropriate environmental, technical, and 
biological data collected from the gear deployments 
were treated as either fixed, categorical, or continuous 
variables. Where there was sufficient replication, these 
variables were considered with the fixed factors of pri- 
mary interest: “method of sorting” (conventional vs. 
water tray); delay in sorting” (2 min vs. 15 min); and 
sampling time (T 0 , T 24 , T 72 or T 120 ); and the random 
factors of fishing days, deployments, and cages in mixed- 
effects logistic models were fitted to the dichotomous 
status (dead vs. alive) of the trawled-and-discarded 
caged school prawns. 
For each experiment, three separate models were 
used for trawled-and-discarded school prawns and com- 
prised data from 1) all deployments, 2) the water-tray 
deployments, and 3) the conventional-tray deployments 
only. A fourth model was fitted to control, caged school 
prawns, and was restricted to the above random factors 
and the fixed effect of sampling time. All models were 
fitted by using the lmer function in the R statistical 
software (The R Foundation for Statistical Computing, 
Vienna, Austria). A stepwise variable search algorithm 
was employed with the most parsimonious fit based on 
the lowest Akaike’s information criterion. The total 
mortalities of school prawns subjected to the four han- 
dling treatments of interest were eventually adjusted 
for deaths to the controls. 
A balanced four-factor analysis of variance (ANOVA) 
was used to examine differences among treatment and 
control groups for the levels of L -lactate (pmol/g) in 
school prawns from one fishing day in the Clarence 
River experiment. The model used the following fac- 
